-
Notifications
You must be signed in to change notification settings - Fork 510
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
5.4 kernel warning: general protection fault with bpf_probe_read
#1435
Comments
Here's the actual boot log, sorry it wasn't available when I first made the bug report. Also, while the panic happens, the host actually keeps working.. its possible that our other nodes do this, and we just hadn't noticed.
|
I've shared this report with our kernel team. Are you using one of the BPF CNI providers like Calico or Cilium? Regarding the |
We are using the Calico CNI in eBPF mode along with the AWS VPC CNI networking setup..
We did not add anything that I knw of that would cause this.. |
Thanks, that should help us repro & verify a fix once we understand the issue. |
For what its worth - turning off eBPF does not disable this warning message. I am not sure why. |
Hi @diranged, I have a quick update to share. We're still working on this internally with our kernel team. In the meantime, we're releasing a Kubernetes 1.20 variant with our v1.1.0 release that uses the 5.10 kernel, where we believe this issue is fixed. If you're able to upgrade, we'd recommend that as the best option. If you wish to continue using your current variant, you can also create a custom build that swaps out the 5.4 kernel for the 5.10 kernel in the variant of your choice. Here’s an example of how to achieve that: 906df655. |
I saw the 1.1.0 release.. looks like we need to wait for EKS 1.20 (coming soon it seems) right? We have to upgrade the cluster before we can upgrade our nodes, I believe. |
EKS has now released 1.20 support! 🎉 |
👍 we're in the upgrade process now.. and i've tested the 1.1.1 release w/ eks 1.20 and no longer see the kernel issue on-startup! yay! |
Hi @diranged, I'm glad to hear that you're no longer running into the issue. Although this is fixed in variants with the 5.10 kernel, other variants with the 5.4 kernel will still have the issue. So I would like to edit this issue to track the fix for our other 5.4 kernel variants. Feel free to unsubscribe from the issue notification if it gets too noisy. |
bpf_probe_read
We have since moved past this kernel version and don't believe it is still an issue. Closing, but feel free to reopen a new issue if there is anything else related. |
@etung Edit: all variants with the 5.4 kernel are affected.
Image I'm using:
In our forked repo (with just a change to the CNI networking/instance-type map text file), we built the v1.0.7 tag to try to deal with a possible issue we're seeing with Systemd. The images are not booting though. We booted 3 servers, and all failed with a nearly identical failure:
Git Sha Built: fdb0353
What I expected to happen:
Ideally it would have booted. :)
What actually happened:
All 3 servers booted and failed like this:
The text was updated successfully, but these errors were encountered: